AI-generated ads achieve parity but lack emotional pull

Ipsos and Syracuse University's Newhouse School found that AI-generated ads achieve surface-level parity with human-made ads but underperform on emotional engagement. In a test of 20 ads, AI outputs met programmatic criteria but scored lower on metrics tied to long-term brand health, according to Ipsos creative excellence lead Daren Poole.

AI-generated ads achieve parity but lack emotional pull Daren Poole, creative excellence lead at Ipsos, writes in Mumbrella that generative AI video tools now produce ads that most viewers cannot reliably distinguish from human-made work. Poole summarises an experiment conducted by Ipsos with Syracuse University's Newhouse School that reverse-engineered a recent Fiat commercial and tasked an AI video model to recreate it. Per the article, the AI-produced spot looked polished but "flatlined" on emotional engagement in testing. Ipsos evaluated 20 ads, half human-produced and half AI-generated, using its sales-validated Creative Spark platform and the Misfits Mindset framework; Poole reports AI outputs met surface requirements but underperformed on measures linked to long-term brand health. The piece warns this "good enough" parity in surface quality can mask systematic shortfalls in emotional resonance. What happened Daren Poole, creative excellence lead at Ipsos , argues in Mumbrella that generative AI video tools now generate ads that many consumers cannot reliably distinguish from human-made work. Per the article, Ipsos and Syracuse University's Newhouse School conducted an experiment that reverse-engineered a recent Fiat commercial and prompted an AI video model to reproduce the brief. The resulting AI-produced spot was described as polished and credible, but it "flatlined" on emotional engagement during U.S. consumer testing, according to Poole. Ipsos reportedly evaluated 20 ads, evenly split between human-produced and AI-produced creative, using its sales-validated Creative Spark platform and the Misfits Mindset framework; the article reports that AI outputs matched surface programmatic criteria but scored lower on measures tied to long-term brand impact. Editorial analysis - technical context Industry-pattern observations: generative models excel at reproducing visible structure, motifs, and programmatic instructions, but published practitioner analysis and the Ipsos example indicate those strengths do not automatically transfer to reliably evoking nuanced emotional responses. For practitioners, that gap often shows up as high precision on scripted cues yet low uplift on memory, empathy, or distinctiveness metrics that predict sustained sales. Context and significance the Mumbrella piece frames this as a strategic trade-off for marketing teams under pressure to move faster and cut budgets. Advertisers substituting rapid AI iterations for human-led concepting risk producing creative that checks technical boxes without delivering the storytelling qualities that correlate with long-term brand metrics. This observation aligns with broader reporting on generative AI in creative workflows where scale and speed sometimes sacrifice depth of insight. What to watch Observers should track replication of Ipsos-style tests across categories and markets, the emergence of evaluation frameworks that weight emotional resonance more heavily, and vendor features that expose provenance or creative intent. Also monitor whether academic or third-party replications confirm the pattern of parity on surface metrics but systematic underperformance on emotion-linked KPIs. Notes on sourcing All factual claims above are drawn from Daren Poole's article in Mumbrella summarising the Ipsos and Syracuse University experiment and Ipsos' Creative Spark results. Poole's piece is the primary public account of the experiment; the article does not include direct quotes from Fiat or the AI vendors used in the test. Scoring Rationale The story highlights empirical limits of generative AI in advertising that matter to practitioners designing creative evaluation and measurement pipelines. It is noteworthy but not a frontier technical breakthrough. Practice with real Ad Tech data 90 SQL & Python problems · 15 industry datasets Active Search Campaigns by BudgetEasy /problems/sql/active-search-campaigns-by-budget High CPC Clicks & Poor Landing PagesMedium /problems/sql/high-cpc-clicks-poor-landing-page Campaign ROAS by Attribution ModelHard /problems/sql/campaign-roas-by-attribution-model 250 free problems · No credit card See all Ad Tech problems /problems/datasets/adtech